TEMPLATE DESIGN © 2007
www.PosterPresentations.c
om
Impact of Artifact Removal from EEG Signals
on Epileptic Seizure Detection
Partho Prosad, Zarif Ahmed & Shojib Ahmed Refath
Supervisor: Dr. Md. Kafiul Islam
Department of Electrical and Electronic Engineering
Independent University, Bangladesh
Abstract Complex Engineering Problems
Complex Engineering Activities
Proposed Total Budget
Conclusions
 The wavelet denoising we used, was based on trial
and error.
 The dataset we selected all consisted of only
seizure patients and no healthy subjects.
 We did not work much on the classifier. We used
standard classifier via NPR tool.
Future Works
Objectives
• Utilize at least 2 artifact removal techniques
on a dataset of epilepsy patients.
• Analyze the impact of artifact removal on the
accuracy of epileptic seizure detection.
Average Accuracy %
Impact of project outcome on the
environment and sustainability
Limitations
 We have found that artifact removal can improve the
accuracy of seizure detection.
 There were two different methods for signal denoising used
which were wavelet and EMD.
 We quantified the artifact removal.
 In conclusion, we have found that EMD gives higher
accuracy compared to wavelet transform for which
decomposition level depends on user and choice of the right
mother wavelet.
 Deep learning can be applied for better classification.
 ICA, adaptive filtering and other signal processing
technique along with EMD and WT can be applied for
improved artifact removal precision.
 We can apply the same process to improve the
classification results for other neurological diseases.
 We can further improve the EMD algorithm by using
automated IMF tuning.
Project Plan – Gantt Chart
Confusion Matrix & ROC (EMD)
Results - Comparative analysis of
Performance Parameters
Work Plan – RACI Matrix
Proposed System
FYDP
Spring 2024
Task Name
1st
Term 2nd
Term 3rd
Term
Jun-23 Jul-23 Aug-23 Sep-23 Oct-23 Nov-23 Dec-23 Jan-23 Feb-24 Mar-24 Apr-24
May-
14
Prepare Plan
Understanding concepts
of Epileptic Seizure
Research Literature
Problem Finding
Prepare & Submit
Project Proposal
Prepare & Submit 1st
Term Progress Report
Prepare & Present 1st
Term Progress
Selecting Suitable
Artifact removal method
, dataset
Study the Result
Prepare & Submit 2nd
Term Progress Report
Prepare & Present 2nd
Term Progress
Classification
Results on various
criteria
Final Report
Final Presentation
Project Demonstration
Confusion Matrix & ROC (WT Artifact)
Period Assignments
Responsible (R), Accountable (A), Consulted
(C) & Informed (I) - RACI Matrix
Start End Status
Dr. Md.
Kafiul
Islam
Partho
Prosad
Zarif
Ahmed
Sojib
Ahmed
Refath
1st
Term
 Prepare Plan C R R R 01.06.2023 30.06.2023
 Review Literature I R R A 01.07.2023 30.08.2023
 Problem Identification C R R R 16.06.2023 30.08.2023
 Prepare Draft Budget I A R A 01.08.2023 30.08.2023
 Prepare, Submit & Present
(Proposal, Progress
Presentation & Progress
Report)
I R R A 18.08.2023 17.10.2023
2nd
Term
 Project Design
(specify the work)
C R R R 01.10.2023 30.11.2023
 Simulation / Hardware
(specify the work)
C R R R 01.11.2023 30.12.2023
 Prepare, Submit & Present
(Presentation & Progress
Report)
I A A A 01.12.2023 15.01.2024
Final
Term
 Testing prototype C R R R 01.01.2024 25.02.2024
 Result & Analysis C R R R 01.02.2024 30.03.2024
 Prepare, Submit & Present
(Final Report &
Presentation)
I A A A 01.03.2024 30.04.2024
 Prepare Poster & Present
Group Demonstration
I R R R 01.04.2024 15.05.2024
Sl Item Justification Price (BDT)
1.
Matlab
Software
License
For pre-processing raw EEG data and
to apply signal processing techniques
for artifact removal
30,150
Total (BDT) 30,150
In word: Thirty Thousand One Hundred and Fifty Taka Only
Epilepsy is a prevalent neurological condition affecting
millions worldwide. It is characterized by recurrent
seizures which can vary significantly in their clinical
manifestation. Electroencephalography (EEG) plays a
crucial role in epilepsy diagnosis and management by
enabling detection of seizure-related brain activity.
However, EEG signals are often contaminated by various
artifacts from sources like eye blinks, muscle activity and
electrode motion. Such artifacts pose a major challenge for
automated seizure detection algorithms by obscuring the
underlying ictal patterns. In this study, we aimed to
evaluate the impact of different artifact removal
techniques on the performance of computerized epileptic
seizure detection from EEG data. Specifically, we wanted
to explore which technique would yield the optimal signal
quality for enabling accurate identification of seizure
events. We applied two popular artifact removal methods -
wavelet transform and empirical mode decomposition - to
preprocess stationary EEG recordings with simulated
artifacts. Technical measures and clinical detection
accuracy were then used to compare the performance of
each technique. Our findings indicate that empirical mode
decomposition more effectively mitigated artifacts,
achieving a higher signal-to-noise ratio of approximately
6dB compared to -4dB for wavelet transform. Seizure
detection precision and sensitivity were also improved
with empirical mode decomposition preprocessing,
exceeding 80% for most metrics. This research highlights
the significance of thorough artifact removal in facilitating
automated seizure detection from EEG signals. By
minimizing noise from extraneous sources, underlying
ictal patterns can be better characterized, aiding epilepsy
diagnosis and management. With ongoing refinement,
such signal processing techniques show promise for
augmenting clinical decision making and improving care
for people living with this condition.
Confusion Matrix & ROC (WT)
Confusion Matrix & ROC (Clean Signal)
WT Simulation EMD Simulation
Experiment workflow
Wavelet Transform (Decomposition)
Empirical Mode Decomposition (EMD)
Performance Analysis of
Artifact Added Signal
Performance Analysis After
Artifact Removal (WT)
Performance Analysis After
Artifact Removal (EMD)
Confusion Matrix
Confusion Matrix & ROC Curve

Undergraduate Thesis - EEE400 - Final Year Design Project - Poster Presentation.pptx

  • 1.
    TEMPLATE DESIGN ©2007 www.PosterPresentations.c om Impact of Artifact Removal from EEG Signals on Epileptic Seizure Detection Partho Prosad, Zarif Ahmed & Shojib Ahmed Refath Supervisor: Dr. Md. Kafiul Islam Department of Electrical and Electronic Engineering Independent University, Bangladesh Abstract Complex Engineering Problems Complex Engineering Activities Proposed Total Budget Conclusions  The wavelet denoising we used, was based on trial and error.  The dataset we selected all consisted of only seizure patients and no healthy subjects.  We did not work much on the classifier. We used standard classifier via NPR tool. Future Works Objectives • Utilize at least 2 artifact removal techniques on a dataset of epilepsy patients. • Analyze the impact of artifact removal on the accuracy of epileptic seizure detection. Average Accuracy % Impact of project outcome on the environment and sustainability Limitations  We have found that artifact removal can improve the accuracy of seizure detection.  There were two different methods for signal denoising used which were wavelet and EMD.  We quantified the artifact removal.  In conclusion, we have found that EMD gives higher accuracy compared to wavelet transform for which decomposition level depends on user and choice of the right mother wavelet.  Deep learning can be applied for better classification.  ICA, adaptive filtering and other signal processing technique along with EMD and WT can be applied for improved artifact removal precision.  We can apply the same process to improve the classification results for other neurological diseases.  We can further improve the EMD algorithm by using automated IMF tuning. Project Plan – Gantt Chart Confusion Matrix & ROC (EMD) Results - Comparative analysis of Performance Parameters Work Plan – RACI Matrix Proposed System FYDP Spring 2024 Task Name 1st Term 2nd Term 3rd Term Jun-23 Jul-23 Aug-23 Sep-23 Oct-23 Nov-23 Dec-23 Jan-23 Feb-24 Mar-24 Apr-24 May- 14 Prepare Plan Understanding concepts of Epileptic Seizure Research Literature Problem Finding Prepare & Submit Project Proposal Prepare & Submit 1st Term Progress Report Prepare & Present 1st Term Progress Selecting Suitable Artifact removal method , dataset Study the Result Prepare & Submit 2nd Term Progress Report Prepare & Present 2nd Term Progress Classification Results on various criteria Final Report Final Presentation Project Demonstration Confusion Matrix & ROC (WT Artifact) Period Assignments Responsible (R), Accountable (A), Consulted (C) & Informed (I) - RACI Matrix Start End Status Dr. Md. Kafiul Islam Partho Prosad Zarif Ahmed Sojib Ahmed Refath 1st Term  Prepare Plan C R R R 01.06.2023 30.06.2023  Review Literature I R R A 01.07.2023 30.08.2023  Problem Identification C R R R 16.06.2023 30.08.2023  Prepare Draft Budget I A R A 01.08.2023 30.08.2023  Prepare, Submit & Present (Proposal, Progress Presentation & Progress Report) I R R A 18.08.2023 17.10.2023 2nd Term  Project Design (specify the work) C R R R 01.10.2023 30.11.2023  Simulation / Hardware (specify the work) C R R R 01.11.2023 30.12.2023  Prepare, Submit & Present (Presentation & Progress Report) I A A A 01.12.2023 15.01.2024 Final Term  Testing prototype C R R R 01.01.2024 25.02.2024  Result & Analysis C R R R 01.02.2024 30.03.2024  Prepare, Submit & Present (Final Report & Presentation) I A A A 01.03.2024 30.04.2024  Prepare Poster & Present Group Demonstration I R R R 01.04.2024 15.05.2024 Sl Item Justification Price (BDT) 1. Matlab Software License For pre-processing raw EEG data and to apply signal processing techniques for artifact removal 30,150 Total (BDT) 30,150 In word: Thirty Thousand One Hundred and Fifty Taka Only Epilepsy is a prevalent neurological condition affecting millions worldwide. It is characterized by recurrent seizures which can vary significantly in their clinical manifestation. Electroencephalography (EEG) plays a crucial role in epilepsy diagnosis and management by enabling detection of seizure-related brain activity. However, EEG signals are often contaminated by various artifacts from sources like eye blinks, muscle activity and electrode motion. Such artifacts pose a major challenge for automated seizure detection algorithms by obscuring the underlying ictal patterns. In this study, we aimed to evaluate the impact of different artifact removal techniques on the performance of computerized epileptic seizure detection from EEG data. Specifically, we wanted to explore which technique would yield the optimal signal quality for enabling accurate identification of seizure events. We applied two popular artifact removal methods - wavelet transform and empirical mode decomposition - to preprocess stationary EEG recordings with simulated artifacts. Technical measures and clinical detection accuracy were then used to compare the performance of each technique. Our findings indicate that empirical mode decomposition more effectively mitigated artifacts, achieving a higher signal-to-noise ratio of approximately 6dB compared to -4dB for wavelet transform. Seizure detection precision and sensitivity were also improved with empirical mode decomposition preprocessing, exceeding 80% for most metrics. This research highlights the significance of thorough artifact removal in facilitating automated seizure detection from EEG signals. By minimizing noise from extraneous sources, underlying ictal patterns can be better characterized, aiding epilepsy diagnosis and management. With ongoing refinement, such signal processing techniques show promise for augmenting clinical decision making and improving care for people living with this condition. Confusion Matrix & ROC (WT) Confusion Matrix & ROC (Clean Signal) WT Simulation EMD Simulation Experiment workflow Wavelet Transform (Decomposition) Empirical Mode Decomposition (EMD) Performance Analysis of Artifact Added Signal Performance Analysis After Artifact Removal (WT) Performance Analysis After Artifact Removal (EMD) Confusion Matrix Confusion Matrix & ROC Curve